Sikoia

Models do the work that removes the manual bottleneck, extracting and validating income and employment from payslips, bank statements and tax returns, analysing open banking data for affordability, and driving anti money laundering screening.

Against that, the platform's founding proposition is aggregation: a unified data platform consolidating credit bureau, public registry, identity, fraud and screening services from a marketplace of partners into one view, with a decision engine and dashboard on top. Strip the models and that consolidation layer, the partner marketplace and the workflow automation all remain useful, which is the Alloy and Signzy position rather than a pure model business.

The company frames its output as evidence for a lender rather than as a decision, describing structured, explainable and auditable financial data that enables firms to decide, and its chief executive puts the requirement in the buyer's own terms: lenders need stronger affordability evidence and decisions they can clearly explain and defend. Auditability is repeated across products, including for suitability assessments under the consumer outcomes regime.

That is the correct division for a verification supplier and it is the Pave position. What holds it below the top grade is that decision automation with dynamic checklists is also offered, so in some configurations the platform does conclude rather than inform, and no threshold or review point is described for those.

Explainability and auditability are stated as properties of the output rather than as aspirations, repeated across products and tied to a specific need, that lenders must be able to defend a decision after the fact. For verification work that is the right control, since a lender can show which document produced which figure.

The company has also used a neutral third party sandbox environment to test and refine its document processing before release, which is external validation infrastructure rather than self assessment. What is absent is measurement: no extraction accuracy, error rate by document type or validation result is published, and for a product whose entire function is reading numbers correctly that is the figure a model risk function would ask for first.

Named institutional customers carry this rather than volume figures. A United Kingdom challenger bank uses the platform for income and employer verification in mortgage decisioning, a building society automates document handling, income verification and affordability checks with it, a car finance challenger embedded it from launch, and a mortgage software provider managing close to 50 billion pounds in loans and servicing more than 650,000 borrower accounts has integrated it with further phases planned into its broker system.

Customers are quoted directly. Against that, no processing volume, application count or measured outcome is published, total funding is modest at around 8.3 million dollars, and coverage is described only as more than three countries.

No data boundary statement was located. Document intelligence improves with exposure to more payslip layouts, employer formats and statement styles, so a model serving many lenders benefits from every applicant flow it sees, and the institutions on this platform compete directly for the same borrowers in the same market. Nothing states whether extraction models are trained on customer submitted documents, whether a lender can decline to contribute, or how applicant material is separated between institutions.

The privacy position rests on licences rather than assurances, which is the strongest form it can take. Authorisation as an account information service provider means open banking access operates under the statutory consent framework, so the customer authorises each connection and can withdraw it, and the subject is therefore a party to the arrangement rather than an object of it.

Registration as a credit reference provider brings statutory obligations on the accuracy of data held about individuals and on their right to see and challenge it. Held at B because no retention schedule, subprocessor list or deletion commitment was located, and the platform handles payslips, tax returns and bank statements whose handling terms are not described.

No attestation, certification, trust centre or enumerated framework was located. Holding two regulatory authorisations brings supervisory expectations around systems, controls and data handling that a regulator can examine, which is a genuine floor and is not the same as a published assurance artifact. For a platform holding payslips, tax returns and bank statements on behalf of banks and building societies, an independently assessed control set is what those institutions' own reviews will request.

The most precise regulatory disclosure recorded in this index, and unlike its nearest rivals it concerns the vendor's own permissions rather than its customers' obligations. The company publishes that it is authorised and regulated by the conduct regulator as an account information service provider under the named payment services statutory instrument, and separately as a credit reference provider, with a distinct firm reference number given for each so a reader can verify both on the public register.

It then states what it is not, declaring itself a credit broker and not a lender, which forecloses the question this index otherwise has to settle by inference. Sector commitment is evidenced through trade association membership in motor finance, and the regulator's review of historic commission arrangements is named as the market context. Tenth A on this axis.

Document extraction quality varies with what the document looks like, and payslip layouts, employer systems, self employment records and irregular or multiple income streams are exactly where automated reading degrades. The company states its motor finance expansion is aimed at lenders wanting to serve customers with complex financial profiles, which is a genuine inclusion objective and also identifies the population whose documents are hardest to read correctly, so the applicants who stand to gain most are the ones most exposed to a misread.

Credit reference provider status brings statutory accuracy duties that partly answer this, and no extraction accuracy by document type, employer format or income pattern was located, nor any analysis of outcomes across borrower groups.

No commercial guarantee or indemnity was located, and the recourse that exists is statutory and therefore stronger. Registration as a credit reference provider places the company under legal duties on the accuracy of information it holds about individuals and gives those individuals a right to obtain and challenge it, so an applicant misrepresented by an extraction error has an actual mechanism rather than a hope.

The account information service authorisation adds consent and withdrawal rights over the open banking data. Auditable output means a lender can reconstruct which document produced a figure. What is missing is anything describing correction of a specific extraction error or notification of a lender whose decision rested on one.

The data chain is named at its most important points rather than described in categories alone, identifying a major credit bureau, a business information provider, an open banking aggregator and a financial crime screening vendor, with the broader partner marketplace described by function across bureaus, registries, identity, fraud and screening.

That lets a lender see whose data supports a verification, which matters because the vendor is itself a regulated credit reference provider passing information onward. What is not disclosed is the model layer, with no provider named for the document intelligence that reads applicant paperwork, and no subprocessor list or hosting arrangement located.

Both ends of the platform are named and evidenced. Upstream, individual data partners are identified rather than described generically, covering a major credit bureau, a business data provider, an open banking aggregator and a financial crime screening vendor that is itself indexed here, alongside a stated marketplace spanning bureaus, public registries, identity verification, fraud and anti money laundering screening.

Downstream, the platform is embedded in a mortgage software provider serving hundreds of thousands of borrower accounts with extension into its broker system planned, and is available through a technology marketplace used by financial institutions. Access is offered both as interfaces for developers and as a ready to use dashboard, so a small broker and a large lender consume it differently.

No hosting provider, region selection, residency commitment or private deployment option was located. Operating under United Kingdom authorisation with expansion across Europe and beyond means applicant documents, bank data and identity records cross more than one regime, and a regulated lender outsourcing verification is expected to know where that material is processed. Nothing published addresses it.

No pricing, packaging or basis of charge was located. One evaluation route exists that most vendors lack: the platform is listed on a neutral technology marketplace with a digital sandbox, which the company has used to let institutions test its document processing before committing, and that lowers the cost of assessment even though it says nothing about price. Nothing indicates whether charge falls per verification, per application, per data source called or as a platform subscription.

Four distinct lending and verification markets are served with named customers in each, spanning residential mortgages, motor finance, business lending and tenant referencing, and the institution types run from challenger banks and building societies through brokers and loan origination software providers to specialist finance challengers. Expansion into motor finance in 2026 came with membership of the sector's trade association division, which signals commitment beyond a marketing page. The limit is geography, with the footprint concentrated in one market and described only as more than three countries.